1. Field of the Invention
The present invention relates to an image processing apparatus and method thereof, which detect a noise containing part of an image obtained by decoding image data which has undergone frequency transformation and lossy compression coding by, e.g., an MPEG (Moving Picture Experts Group) scheme or the like, and remove that noise part.
2. Description of the Related Art
Conventionally, as image coding schemes, Motion JPEG, Digital Video, and the like as intraframe coding schemes, H.261, H.263, MPEG-1, and MPEG-2 using interframe predicting coding are available. In recent years, a coding scheme such as H.264 or the like is known. These coding schemes are internationally standardized by ISO (International Organization for Standardization) and ITU (International Telecommunication Union).
MPEG-2 as a representative coding scheme is called lossy coding, and decoded data of encoded image data does not become the same as original image data. This is caused by quantization executed after DCT conversion upon coding. These coding schemes reduce the information size by removing redundancy included in an image signal by using the statistical nature of the image signal. That is, since human vision is insensitive to high-frequency components of an image, these high-frequency components are coarsely quantized to remove redundancy, thus obtaining high coding efficiency.
However, since lossy coding that executes DCT conversion of this type compresses data by limiting high-frequency components, noise called mosquito noise is generated at an edge part of an image or around a moving object, thus causing image quality deterioration of the decoded image. Methods of reducing mosquito noise have been proposed. For example, a method of detecting an edge part of an input image signal and applying filtering processing to the edge part of the image signal using a noise removal filter to obtain an image signal from which mosquito noise is reduced has been proposed. Refer to the international standards of ISO/IEC for detailed contents of MPEG and the like.
As a technique for executing such noise removal, Japanese Patent Laid-Open No. 10-13718 describes a technique for detecting horizontal and vertical edges after detection of a motion vector, and then applying filtering for respective dimensions.
However, in the above related art, since filtering processing for mosquito noise removal is applied to the edge part of an image signal irrespective of the presence/absence of mosquito noise in the image signal, an image deteriorates in a part where no mosquito noise is generated. When an image includes a texture area including fine edges, this area is erroneously determined as an edge, and texture information is deleted, resulting in deterioration of image quality.
These problems are posed since the noise removal processing is applied to the edge part of an image signal without accurately detecting a part where noise such as mosquito noise or the like included in an image obtained by decoding image data that has undergone frequency conversion and lossy coding is generated. Also, these problems are posed since the noise removal processing is applied while considering a texture area including fine edges of an image signal as a noise part.
Furthermore, if processing such as IP conversion for converting an interlaced signal into a progressive signal or the like is to be done after the noise removal processing, the noise of the decoded image must be removed so as not to impose any adverse influence on that processing. In other words, noise must be removed while maintaining interlaced information.